Why is this important?
Posted: Thu Feb 06, 2025 9:26 am
Create a hypothesis
Now that you’ve decided what you want to start testing, it’s time to create a hypothesis.
Well, without hypotheses you end up testing random stuff without any rhyme or reason.
Hypothesis helps you test things based on logic and data .
Not intuition
For example, last year I decided to test a scroll box on Backlinko:
Scroll Box
My hypothesis is that scroll boxes can improve morocco mobile database newsletter signups without hurting conversions elsewhere like exit-intent popups.
Before I started testing, I made sure to note this assumption:
Assumptions
How to run an AB test
Here are two tips to help you run a split test:
First, start testing breaking changes.
One of the most common AB testing newbie mistakes is testing tiny variations.
Like button color
Instead, you want to start by testing two very different versions of your page. Then, over time, use smaller changes to optimize the winning page.
For example, when I was testing on the Backlinko homepage last year, I could test my button copy, logo, title font size… and a thousand other little tweaks.
Now that you’ve decided what you want to start testing, it’s time to create a hypothesis.
Well, without hypotheses you end up testing random stuff without any rhyme or reason.
Hypothesis helps you test things based on logic and data .
Not intuition
For example, last year I decided to test a scroll box on Backlinko:
Scroll Box
My hypothesis is that scroll boxes can improve morocco mobile database newsletter signups without hurting conversions elsewhere like exit-intent popups.
Before I started testing, I made sure to note this assumption:
Assumptions
How to run an AB test
Here are two tips to help you run a split test:
First, start testing breaking changes.
One of the most common AB testing newbie mistakes is testing tiny variations.
Like button color
Instead, you want to start by testing two very different versions of your page. Then, over time, use smaller changes to optimize the winning page.
For example, when I was testing on the Backlinko homepage last year, I could test my button copy, logo, title font size… and a thousand other little tweaks.